At a chemical synapse, the arrival of an action potential (AP) at the nerve terminal activates voltage-gated calcium channels (VGCCs), thereby inducing Ca²⁺ influx into the terminal. Consequently, the fusion of synaptic vesicles (SVs) into the plasma membrane is triggered and neurotransmitters are released. VGCCs are composed of a pore-forming α1 subunit that is encoded by 10 genes and classified into three subgroups characterized with electrophysiological and pharmacological properties: Ca_V1 (L-type), Ca_V2 (N, P/Q, and R-type), and Ca_V3 (T-type) channels. The Ca_V2 channels [Ca_V2.1 (P/Q-type), Ca_V2.2 (N-type), and Ca_V2.3 (R-type)] are highly expressed in chemical synapses of the mammalian central nervous system (CNS) and are concentrated at the presynaptic active zone (AZ). The AZ consists of protein complexes of various molecules including target soluble N-ethylmaleimide-sensitive factor attached protein receptor (t-SNARE) proteins (syntaxin and SNAP-25) and scaffold proteins (e.g., RIMs, Munc-13, Bassoon, Piccolo, and ELKS) as well as VGCCs (Figure 1A). Fast and precise synaptic transmission is accomplished through the coupling of Ca_V2 channels and the Ca²⁺-sensor proteins (e.g., synaptotagmins) of release-ready docked SVs, which are adjacent to each other (Figure 1B). The issue of the required number of Ca_V2 channels for the fusion of a single SV is still under debate. Previous studies using squid giant synapses and chick ciliary ganglion calyx synapses have proposed that Ca²⁺ influx through a single VGCC can generate the fusion of an SV if the Ca²⁺ sensor is sufficiently close to the channel (Stanley, 2016). In chemical synapses of the mammalian CNS, however, Ca_V2 channels make clusters in the AZs, indicating that the Ca²⁺ influx through multiple Ca_V2 channels in a cluster generates the fusion of an SV.

In this study, we have highlighted recent progress toward understanding the distribution and functions of Ca_V2 channels, mainly focusing on the findings based on an immunoelectron microscopic technique called sodium dodecyl sulfate-digested freeze-fracture replica labeling (SDS-FRL). SDS-FRL can be used to visualize the two-dimensional distribution of membrane molecules with high sensitivity and resolution (Masugi-Tokita et al., 2007) and is suitable for quantitative analysis of the number, density, and clustering of Ca_V2 channels in the AZs. Among the three Ca_V2 subunits, Ca_V2.1 has been most extensively examined by several groups and, thus, is mainly discussed here. We have also described the technical notes of SDS-FRL for accurate cluster analysis and have discussed the functional roles of Ca_V2 channel clustering on neurotransmitter release with different topographical models of Ca_V2 channels and SVs.

Synaptic vesicles are thought to be docked near VGCCs in the AZs for the efficient triggering of their exocytosis in response to an AP. The geometric differences of Ca_V2 channels, such as their clustering and the distance from docked SVs, characterize the variety of the presynaptic properties. However, due to the high-density protein accumulation in the AZs, the accessibility of antibodies against Ca_V2 subtypes is potentially attenuated in the conventional immunofluorescence or pre-embedding immunoelectron microscopy. In contrast, the accessibility of the antibody to Ca_V2 channels is improved in SDS-FRL because the epitopes are exposed on the two-dimensional fractured face of the membrane after most of the cytosolic proteins that could mask the epitopes are washed out during the SDS treatment. However, SDS-FRL relies on random fracturing of frozen slices, making the identification of labeled profiles more difficult compared with volume data obtained by the three-dimensional reconstruction of serial sections. SDS-FRL has been utilized for quantitative analysis of the nanoscale distribution of Ca_V2 channels at the AZs in rodent brains (Holderith et al., 2012; Indriati et al., 2013; Baur et al., 2015; Nakamura et al., 2015; Éltes et al., 2017; Kusch et al., 2018; Luján et al., 2018a,b; Rebola et al., 2019; Eguchi et al., 2020; Kleindienst et al., 2020; Bhandari et al., 2021). Figure 1C shows the examples of the replica labeling for Ca_V2.1, 2.2, and 2.3 on presynaptic boutons in different synapses (Éltes et al., 2017; Bhandari et al., 2021). The quantitative analysis of the gold particle distribution suggests that all subtypes of Ca_V2 channels are accumulated in the AZs of presynaptic boutons. However, Cav2.3 has a high density in the surroundings of AZ too, which gradually decreases with distance from the AZ (Bhandari et al., 2021), whereas Cav2.1 shows very low density in the extra-AZ areas (Figure 1D).

To estimate the absolute number of Ca_V2 channels, a one-to-one relationship between single channels and gold particles in SDS-FRL would be ideal (labeling efficiency = 100%). However, the immunogold labeling efficiency in replica samples can be attenuated or amplified for various reasons. Underestimation could occur because not all channel proteins are allocated to the protoplasmic (P-) or exoplasmic (E-) faces for the detection by primary antibodies against the intracellular or extracellular epitopes, respectively. Even if all epitopes are present on the P- or E-faces, not all may be bound to primary antibodies because of masking by associated proteins or low avidity of the antibodies. The immunolabeling conditions including antibody concentrations, temperature of SDS treatment, and contents of blocking and antibody incubation solutions should be optimized too (Kaufmann et al., 2021). In contrast, overestimation could occur because of the aggregation of primary or secondary antibodies or multiple binding of these antibodies to a single target (Figure 2A). In our previous studies, we calibrated the labeling efficiency of SDS-FRL with a Ca_V2.1 antibody in the cerebellum (Indriati et al., 2013; Miki et al., 2017) and calyx of Held (Nakamura et al., 2015), using the number of functional P/Q-type channels deduced by electrophysiological and Ca imaging data. However, some parameters used in the calibration were obtained in different species, ages, or brain regions. In addition, Ca²⁺ channels in the extra-AZ area were neglected. In this study, to estimate a more accurate labeling efficiency, we used Ca_V2.1 labeling in mossy fiber boutons (MFBs) of dentate gyrus granule cells (GCs) innerving CA3 pyramidal neurons (PNs) in the rat hippocampus. Previous electrophysiological and ultrastructural studies have reported the total number of Ca_V2.1 channels to be 1,320 per single MFB in rats at postnatal day (P) 21–22 (Li et al., 2007), and the total surface area of all AZs (∼30 AZs in an MFB) and extra-AZ as 3.33 and 75.5 μm², respectively, in a single MFB of P28 rat (Rollenhagen et al., 2007). Using P28 rats, we labeled Ca_V2.1 in MFBs using an anti-Ca_V2.1 antibody (Frontier Institute, Supplementary Table 1 and Figure 1C), which is essentially the same as those used in our previous studies (Indriati et al., 2013; Nakamura et al., 2015; Luján et al., 2018a,b; Eguchi et al., 2020). The AZs in the MFBs were automatically demarcated using a deep learning-assisted software (Darea) with manual correction (Kleindienst et al., 2020) based on higher densities of intramembrane particles (IMPs) than extra-AZ on the P-face (Hagiwara et al., 2005; Figure 1D). The mean area of the complete AZs in the MFBs in replicas was ∼0.1 μm², which was close to the values obtained by the three-dimensional reconstructions of the MFBs (0.09–0.13 μm², Rollenhagen et al., 2007). The densities of gold particles labeling Ca_V2.1 in the AZs and extra-AZs were 232 and 0.84 particles/μm², respectively. This result and the area of AZ/extra-AZ suggest that 92% of Ca_V2.1 channels are present in the AZs. Thus, the density of Ca_V2.1 at AZs is estimated to be (1,320 channels × 0.92)/3.33 μm² = 365 channels/μm², and the Ca_V2.1 labeling efficiency is calculated to be 64%. This value is very close to previously reported estimates using the same anti-Ca_V2.1 antibody at rat cerebellar parallel fiber (PF)-Purkinje cell (PC) synapses (63%, Indriati et al., 2013; 62%, Miki et al., 2017) and at rat calyx of Held synapses (62%, Nakamura et al., 2015). Using the same Ca_V2.1 antibody and similar labeling conditions, we compared the estimated number and density of Ca_V2.1 channels in different types of CNS synapses (Supplementary Table 2). In the hippocampus, three types of excitatory synapses showed a range of 9.8–24.2 channels as per AZ but with similar densities (288–358 channels/μm²) and nearest neighbor distances (NND, an index for local density, 27–29 nm), indicating that the difference in number is mostly ascribable to that in the AZ size. Interestingly, the center-periphery index (CPI, Kleindienst et al., 2020) showed a significantly lower value (P < 0.0001) in perforant path-to-granule cell (PP-GC) synapses (0.42) than that in MFB synapses (0.67), indicating that Ca_V2.1 is localized nearer to the AZ center in PP-GC synapses (Figures 1D,E). In the cerebellum, the Ca_V2.1 density in PF boutons was higher in synapses made on molecular layer interneurons (MLI, 827 channels/μm²) than those on PCs (416 channels/μm²), indicating target cell-dependent regulation of Ca_V2.1 density. Although the Cav2.1 numbers are linearly correlated with release probability in excitatory synapses of CA3 hippocampal pyramidal cells (Holderith et al., 2012), the number or density does not necessarily correlate with release probability or Ca²⁺ influx across different types of synapses (Éltes et al., 2017; Rebola et al., 2019). Functional properties of single channels or spatial relationship between Ca²⁺ channel clusters and release-ready docked SVs should also affect the synaptic efficacy and temporal precision of neurotransmission.

Previous studies have demonstrated not only the number/density of Ca_V2.1 channels in the AZs but also their clustering (Indriati et al., 2013; Nakamura et al., 2015; Miki et al., 2017; Rebola et al., 2019; Kleindienst et al., 2020), indicating the massive influx of Ca²⁺ through the clustered channels to effectively increase the Ca²⁺ level at the nano-spots in the AZs. To estimate clustering patterns accurately, a high labeling efficiency as described earlier is desirable. Moreover, some technical pitfalls causing artificial gold particle clustering must be excluded (Figure 2A). A valuable indicator of the artificial cluster formation is the non-specific background labeling of gold particles on the “wrong” face or cross-fracture (Figure 2B). In the case of our Ca_V2.1 labeling with the antibody against an intracellular epitope, gold particles on the E-face could serve as background labeling. Figure 2B shows examples of such non-specific labeling with different anti-Ca_V2.1 antibodies on the dendritic E-face in mouse cerebellar PCs. The gold particles with antibody A were clustered with 4–6 particles, whereas those with antibody B were mostly isolated single particles. If the background gold particles make clusters, the results of clustering analysis for the specific labeling could be misleading. Possible causes of this artificial clustering can be examined and eliminated as follows (Figure 2A).

A great variety of methods have been developed to detect particle clusters by comparison with random distributions; Szoboszlay et al. (2017) evaluated the benefits and effectiveness of the implemented methods, and they recommended the comparison of the mean NND between particles as a method to investigate whether gold particles form uniform or clustered patterns compared with random point patterns. If the particles are clustered, the mean NND will be significantly smaller than that of randomly distributed particles computed from Monte Carlo simulations. They also recommended the spatial autocorrelation function (ACF), described as g(r), a derivative of Ripley’s K function based on all-to-all distances between particles, as another useful method to detect clustering. g(r) values close to 1 indicate the random distribution of particles, values < 1 indicate a uniform distribution, and values > 1 indicate clustering. In Figure 2E, we analyzed the Ca_V2.1 clustering in the AZs at PF-PC synapses of the mouse cerebellum. We calculated the mean values of NND and g(r) of the real particles distributed within the AZs on the replicas and the simulated particles with random distribution (200 times simulations for each examined AZ). The mean values of real NND and g(r) were significantly smaller and larger, respectively, than their corresponding simulated values (P < 0.05, paired t-test, n = 20 AZs), suggesting that Ca_V2.1 is distributed in a clustered manner within these AZs.

Several methods have been utilized to automatically detect each cluster of gold particles on the replica. A simple way to define a cluster is to consider particles that exist within a radius (e.g., 100 nm) of each particle to compose a cluster (Nakamura et al., 2015). An advanced method is density-based spatial clustering of applications with noise (DBSCAN), which is a density-based clustering algorithm that groups together points that are tightly clustered while marking points that are less dense and singly distributed as outliers (Szoboszlay et al., 2017). Figure 2E demonstrates the cluster analysis of Ca_V2.1 immunolabeled AZs using DBSCAN with two parameters: the radius of a neighborhood (i.e., the maximum distance between two particles for one to be considered as in the neighborhood of the other) as mean NND plus two times its SD and the minimum number of particles necessary to form a cluster as 3. The DBSCAN algorithm detected 2.6 ± 0.2 clusters in PF-PC synapses (n = 65 AZs). The gold particle clusters consist of 6.8 ± 0.3 particles (n = 167 clusters), indicating that 10.6 ± 0.5 channels form a cluster in the AZs estimated with the labeling efficiency (64%).

Supplementary Table 2 shows the summary of the Ca_V2.1 channel clustering, which was detected in all types of rodent CNS synapses analyzed so far in the previous and present studies using the same primary antibody (refer to the section “Materials and Methods” and Supplementary Table 1). Although the number and density of clusters vary depending on the synapse types, the mean NNDs and numbers of particles per cluster were quite similar, indicating a common mechanism of Ca_V2.1 cluster formation. The distribution of Ca_V2.1 clusters should be examined relative to docked SVs distribution for considering their functional implications (Nakamura et al., 2015). Several different models of the topographical relationship between Ca_V2 channel clusters and docked SVs have been proposed as discussed in the following section (Figure 2F).

Voltage-gated calcium channels in the AZs are coupled with calcium sensors, synaptotagmins, on the docked SVs to efficiently evoke neurotransmitter release. Although it is still being debated about how many VGCCs are required to evoke a vesicle fusion, SDS-FRL observations have provided crucial information to understand the contribution of VGCC distribution to neurotransmitter release. At calyx of Held synapses in the rat auditory brainstem, the number of Ca_V2.1 in clusters increased during hearing onset (P7 vs. P14) with no changes in NND, and the cluster area was expanded (Nakamura et al., 2015). This developmental change results in tighter coupling of Ca_V2.1 channel clusters with vesicle fusion sites. A computational simulation based on this as well as electrophysiological and Ca²⁺ imaging results predicts that multiple docked SVs are located at the perimeter of a Ca_V2.1 cluster, and the Ca_V2.1 number in a cluster mainly determines the vesicular release probability (“perimeter release model,” Figure 2F). In contrast, at PF-MLI synapses of the mouse cerebellar cortex, NNDs between particles were significantly shortened during postnatal development (postnatal week 2 vs. 4) without changes of Ca_V2.1 number in the cluster (Miki et al., 2017). This tighter VGCC arrangement may allow for more efficient coupling between Ca²⁺ entry and SV release. Interestingly, the numbers of SV docking sites (estimated from electrophysiological experiments) and Ca_V2.1 clusters showed a close correspondence, leading to propose “one-to-one stoichiometry model” (Figure 2F). In contrast, Rebola et al. (2019) proposed that Ca_V2.1 channels were not clustered but simply excluded from a 50 nm zone around docked SVs (“exclusion zone model,” Figure 2F) at cerebellar PF-PC synapses. However, this contrasts with observations in GABAergic stellate cell synapses, where SVs are tightly associated (∼10 nm) with the perimeter of VGCC clusters consistent with the perimeter release model (Rebola et al., 2019). In the exclusion zone model, the release probability is determined by the radius of the exclusion zones rather than the number of Ca_V2.1 in the AZs.

Sodium dodecyl sulfate-digested FRL studies demonstrate the nanoscale distribution of Ca_V2 channels at AZs of presynaptic terminals. Although a single VGCC can generate the fusion of a single SV (Stanley, 2016), Ca_V2 channels form clusters (approximately 10 channels) in the AZs of many synapse types in the CNS. The distribution of Ca_V2 channels, especially topographical relationships between Ca_V2 channel clusters and Ca²⁺ sensors of docked SVs, and the number of channels in a cluster modulate the properties of neurotransmitter release. The dynamic changes in the Ca_V2 channel clustering might contribute to the presynaptic plasticity of synaptic transmission. It remains unclear how the distribution of Ca_V2 channels and their topographical relationships with SVs are regulated in different types of synapses. The quantitative nanoscale analysis of Ca_V2 channel distribution by SDS-FRL, combined with simultaneous visualization of the release-ready docked SVs, should be useful to unveil these questions.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by the Ethics Committee of IST Austria.

KE and RS conceived and wrote the manuscript. All authors conducted electron microscopic experiments and contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.